RevBrain

Inspiration

Sales teams keep numbers in a database and reasons in notes. To answer "which of my accounts are at risk, and why?" you have to look in both places. I wanted one agent that does it for you.

What it does

Traceview is a company brain for sales teams. You ask a question in plain English and it answers with the real number from the database, the reason from account notes stored in HydraDB, and what to do next from sales playbooks stored in HydraDB.

It also remembers each person's preferences between conversations, and shows how accounts, people and contracts are connected.

How I built it

Traceview started as an earlier project of mine that answered questions with SQL and local search. For this hackathon I added HydraDB as its brain: shared knowledge, per-user memory, and a graph of accounts, people and contracts.

It uses FastAPI and Postgres on the backend, React on the frontend, and OpenAI for the answers. I built it with Claude Code.

Challenges

In testing, general playbooks ranked above an account's own note, so the account's notes are now searched separately and shown first.

In my tests, HydraDB returned graph results only in its slower "thinking" mode, so account questions use that mode.

Saving every conversation as memory was too noisy, so it now saves only facts about the user.

What I learned

A company brain depends on knowing who is asking and which account the question is about.

Built With

• fastapi
• hydradb
• openai
• pgvector
• postgresql
• python
• react
• vite